Optimization method of heat supply system

Abstract

The invention provides an optimization method of a heat supply system, which comprises the following steps: acquiring a real-time heat source operation state, a power grid load, a user demand and an environment constraint condition of the heat supply system, extracting characteristic information of sudden cooling and equipment fault information, and determining a heat source priority and a response speed initial value; acquiring parameters required by a cost function corresponding to the adjusted heat source priority sequence, determining cost weights of different heat sources after adjustment through a linear weighting method, and outputting a cost optimization sequence; identifying a power grid load according to the cost optimization sorting and the environment constraint condition, and if the power grid load is lower than a preset threshold value, calculating a heat source output distribution initial scheme through a dynamic programming algorithm to obtain a preliminary output matrix; and according to the optimized output distribution scheme, summarizing the total heat source emission amount, and if the total heat source emission amount exceeds a preset threshold value, reducing the high-emission heat source output proportion through an iterative adjustment algorithm, and outputting a target output distribution matrix.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an optimization method for a heating system. Background Art

[0002] Multi-heat-source combined heating systems play a crucial role in the modern energy sector. Their core objective is to achieve efficient, stable, and low-carbon heating by integrating multiple heat sources, directly related to the balance between energy efficiency, environmental protection, and economic benefits. With the acceleration of urbanization and the transformation of energy structures, optimizing the scheduling strategies of multi-heat-source systems has become a key issue in enhancing the resilience and sustainability of energy systems. However, existing methods often exhibit significant limitations when dealing with dynamic changes in complex scenarios. Traditional scheduling strategies often rely on single-source priority or static cost optimization, making them difficult to adapt to variable operating conditions such as sudden cooling, equipment failures, or grid load fluctuations. In particular, they lack flexibility and systematicity in integrating multi-source coordination and user demand response, resulting in reduced heating efficiency or increased operating costs. The core challenge in this area lies in how to coordinate the relationship between heat source priority, response speed, and cost functions in complex multi-source systems, while also considering environmental constraints and grid interaction properties. Taking gas-fired combined cycle cogeneration units, gas-fired boilers, industrial waste heat recovery systems, and power plant waste heat as examples, when industrial waste heat is interrupted due to equipment failure or the power grid is in a low-load situation, the output distribution among the heat sources needs to be adjusted quickly, but existing models often cannot effectively balance the contradiction between rapid response and economy. In addition, after introducing user-side demand response as a variable, the complexity of the system is further increased. How to achieve global optimization in the multi-party interest game has become a technical problem that needs to be broken through. These unresolved factors directly lead to the difficulty in ensuring the stability and efficiency of the heating system in extreme scenarios. Therefore, how to construct a multi-heat source output intelligent scheduling model based on game theory in the scenario of sudden cooling, industrial waste heat interruption, and low-load of the power grid, while incorporating user-side demand response into the game framework to optimize the output distribution of the heating system, has become a key issue that needs to be solved in this study. Summary of the Invention

[0003] The present invention provides a method for optimizing a heating system, which mainly includes: Obtain the real-time heat source operating status, grid load, user demand, and environmental constraints of the heating system, extract characteristic information of sudden cooling and equipment failure information, and determine the heat source priority and initial response speed value; The interruption ratio is calculated based on sudden cooling characteristics and equipment failure data. If the interruption ratio exceeds the preset threshold, the heat source output limit is calculated based on historical data and the current load valley state, and the heat source priority sequence is adjusted. Obtain the parameters required for the cost function corresponding to the adjusted heat source priority sequence, determine the cost weights of different heat sources after adjustment through the linear weighting method, and output the cost optimization sorting; Identify the grid load according to the cost optimization sorting and environmental constraint conditions. If the grid load is lower than the preset threshold, calculate the initial scheme of heat source output distribution through the dynamic programming algorithm to obtain the preliminary output matrix; Obtain the grid interaction attributes, combine the preliminary output matrix with the user-side demand response data, extract the fluctuation trend of the peak-valley characteristics from the real-time heat consumption changes, determine the short-term prediction value of the user demand through time series analysis, and output the demand response adjustment coefficient; Combine the preliminary output matrix with the demand response adjustment coefficient, construct a game model, extract the game parameters from the heat source priority, cost function and user demand, calculate the output game result between heat sources, and obtain the optimized output distribution scheme; According to the optimized output distribution scheme, summarize the total heat source emissions. If the total emissions exceed the preset threshold, reduce the output proportion of high-emission heat sources through the iterative adjustment algorithm, and output the target output distribution matrix; Identify the grid interaction attributes corresponding to the output distribution matrix, identify the heat source output volatility. If the volatility is lower than the preset threshold, transmit the output distribution matrix to the heat source control system to determine the heat source output scheduling instruction.

[0004] Further, obtain the real-time heat source operation status, grid load, user demand, and environmental constraint conditions of the heating system, extract the characteristic information of sudden temperature drops and equipment failure information, and determine the initial values of heat source priority and response speed, including: collecting the real-time heat source operation status data set from the heating system monitoring device, collecting the load data set from the grid monitoring unit, collecting the total heat consumption data set through the intelligent meter at the user end, collecting the carbon emission concentration data set and pollutant concentration data set from the environmental monitoring device, and collecting the operation status data set of the heating equipment in real time. Extract the temperature monitoring data from the heating system monitoring device, calculate the temperature change rate of the temperature data at multiple monitoring points using the central difference method, and judge the sudden temperature drop event by setting the temperature change rate threshold to obtain the temperature drop amplitude data set and temperature drop rate data set. Use the operation status data set of the heating equipment for data preprocessing, and obtain the standardized equipment operation data through data standardization processing. If there are abnormal fluctuations in the standardized equipment operation data, identify the fault type based on the pre-trained deep neural network. Calculate the fault response duration data according to the fault type and the standardized equipment operation data, and quantitatively evaluate the fault severity through the fault diagnosis rule base to obtain the fault severity level data. Use the temperature drop amplitude data set, temperature drop rate data set, and fault severity level data to establish a heat source response model, and obtain the heat source supply priority data and heat source response speed data through model calculation. Allocate the response speed index to each heat source unit according to the heat source supply priority data, and dynamically adjust the heat source response speed data using the response speed compensation algorithm to generate the heat source priority data set and the initial response speed value data set.

[0005] Further, according to the sudden cooling characteristic information and equipment failure data, calculate the interruption ratio. If the interruption ratio exceeds the preset threshold, calculate the upper limit of heat source output through historical data and the current low-load state, and adjust the heat source priority sequence, including: set the heat supply interruption weight coefficient matrix according to the sudden cooling amplitude data and the fault duration data, obtain the initial value of the heat supply area interruption ratio through matrix calculation, use the time-series cumulative calculator to record the heat supply interruption ratio data at each moment, and read the interruption ratio threshold data from the preset parameter library. Obtain the load fluctuation data from the heat supply load acquisition device, set the sliding time window of the load fluctuation data to 24 hours, use the exponentially weighted moving average method to obtain the smoothed load curve, and use the density clustering algorithm to divide the load curve into time periods to obtain the load valley period data. Extract the load records of the past 30 days from the historical database, perform normalization processing on the recorded data to obtain the normalized load data, calculate the load change trend curve through support vector regression, and calculate the heat source operation interval in combination with the current load value data and the load valley period data. Establish a heat source output limit matrix using the heat source operation interval data, obtain the upper limit value of the heat source adjustment rate from the preset parameter library, and calculate the adjustment rate correction coefficient based on the heat source response time curve. If the initial value of the heat supply interruption ratio exceeds the interruption ratio threshold data, calculate the upper limit value of the heat source output according to the heat source output limit matrix and the adjustment rate correction coefficient, and generate a priority dynamic adjustment sequence in combination with the heat source priority reference data.

[0006] Further, obtain the parameters required for the cost function corresponding to the adjusted heat source priority sequence, determine the cost weights of different heat sources after adjustment through the linear weighting method, and output the cost optimization sorting, including: read the heat source priority number and the corresponding fuel unit price value from the heat source database, collect the equipment operation efficiency data through the heat source operation monitoring device, obtain the emission factor data from the on-line pollutant monitor, and construct a cost parameter data set. Calculate the fuel consumption per unit time data according to the heat source type in the cost parameter data set, calculate the cumulative operation duration data based on the equipment operation state parameters, and obtain the standard emission concentration data using the standardized unit conversion method. Use the multiple linear regression method to calculate the initial cost weight coefficient, optimize the weight coefficient through regression residual analysis, and generate the standardized cost weight coefficient. Use the standardized cost weight coefficient to perform weighted combination on the fuel consumption per unit time data, the standard emission concentration data, and the cumulative operation duration data to obtain the comprehensive cost function value. Establish a normalization processing matrix for the comprehensive cost function value, perform linear weighting calculation on the normalized cost data in combination with the actual heat source output data to generate the cost optimization sequence data. Establish a heat source operation priority lookup table based on the cost optimization sequence data, update the heat source scheduling order data through the priority mapping relationship, and output the heat source cost optimization sorting result.

[0007] Furthermore, the grid load is identified according to cost optimization sorting and environmental constraint conditions. If the grid load is lower than the preset threshold, the initial scheme of heat source output allocation is calculated by the dynamic programming algorithm to obtain a preliminary output matrix, including: reading the grid load value and load fluctuation value from the grid monitoring device, obtaining the optimization sorting number according to the cost optimization sorting data, collecting pollutant concentration data from the environmental monitoring device, and constructing a load evaluation data set. The load evaluation data set is normalized by a data filter, the load fluctuation trend is predicted by support vector regression, and the environmental constraint index is calculated based on the pollutant concentration data. A recursive neural network is used to perform matching calculations on the heat source capacity value and the heating load amount, the adjustment speed value and the allocation weight value are obtained from the heat source parameter library, and a constraint boundary is established for the heat source output value. If the grid load value is lower than the load threshold, the dynamic planner is used to optimize the calculation of the heat source output, and the preliminary output matrix value is generated according to the constraint boundary and the environmental constraint index.

[0008] Furthermore, the grid interaction attributes are obtained, combined with the preliminary output matrix and the user-side demand response data, the fluctuation trend of the peak-valley characteristics is extracted from the real-time heat consumption change, and the short-term prediction value of the user demand is determined through time series analysis, and the demand response adjustment coefficient is output, including: obtaining the time-of-use electricity price data and the grid frequency data from the grid management center, calculating the grid interaction index according to the preliminary output matrix and the time-of-use electricity price data, collecting the real-time heat consumption data and the user room temperature data, and establishing a user load data set. The user load data set is reconstructed according to the time series, the heat consumption fluctuation characteristics are extracted by using a recursive neural network, the peak-valley load interval is identified according to the time period division mark, and a heat consumption fluctuation curve is generated. The user room temperature data and the time-of-use electricity price data are normalized by the data standardization method, the user response characteristic value is calculated in combination with the heat consumption fluctuation curve, and a user response data set is established. The long short-term memory network is used for time series prediction of the user response data set, the prediction interval is set by a sliding time window, and the short-term prediction data of the user demand is obtained. The response characteristic matrix is constructed according to the peak-valley load interval and the short-term prediction data of the user demand, the user response mode is extracted by using the matrix decomposition method, and the user response reference value is calculated. A response adjustment function is established based on the user response reference value and the grid interaction index, and the demand response adjustment coefficient is obtained through function calculation.

[0009] Furthermore, by combining the preliminary output matrix with the demand response adjustment coefficient, a game model is constructed. Game parameters are extracted from the heat source priority, cost function, and user demand. The output game results among heat sources are calculated to obtain an optimized output allocation scheme, including: extracting the heat source output reference data from the preliminary output matrix, obtaining the priority weight data from the priority sequence, constructing an initial game revenue matrix using the cost function value and user demand, and establishing game constraint conditions for the demand response adjustment coefficient. The upper limit value of heating capacity and the response rate limit value are extracted from the heat source operation database, and a game parameter set is established in combination with the game constraint conditions. The game parameter set is normalized to obtain standardized game parameters. Deep reinforcement learning is used to train the standardized game parameters, and the revenue of each heat source's competition strategy is calculated through the reward function to generate a heat source game strategy dataset. A heat source competition game matrix is constructed based on the heat source game strategy dataset, and a Nash equilibrium solver is used to calculate the game equilibrium solution to obtain the heat source output equilibrium scheme. An adaptive neural network is used to dynamically adjust the heat source output equilibrium scheme, and operation constraints are verified based on the upper limit value of heating capacity and the response rate limit value. The actual output data of each heat source is extracted from the operation constraint verification results, and the multi-objective programming method is used to optimize the configuration of the actual output data to generate an optimized output allocation scheme.

[0010] Furthermore, obtain the heat source priority data, extract game parameters from the data. The game parameters include cost and priority, calculate the output weight of each heat source in combination with the cost function, determine the output distribution, analyze the output relationship among heat sources. If the output of a certain heat source exceeds the preset threshold, then adjust the weight of this heat source to obtain the corrected output distribution, calculate the game result, determine the output balance state among heat sources, and obtain the output scheme by adjusting according to the priority through the sorting method, including: reading the heat source priority number and operation cost data from the database, using the parameter extractor to separate the game parameter set, calculating the initial weight coefficient of the heat source according to the cost function, and generating the heat source priority sorting table. Based on the heat source priority sorting table, construct the initial heat source output distribution matrix, obtain the upper limit data of the response rate from the heat source operation parameter library, and use the recursive neural network to predict the heat source output distribution data for each time period. Compare the heat source output distribution data with the threshold. If the output value exceeds the preset output threshold, then use the linear weighted calculator to reallocate the weight coefficient to obtain the corrected heat source output data. The adaptive compensation algorithm is used to optimize the corrected heat source output data, and the constraint verification is carried out through the upper limit data of the response rate to generate the optimized heat source output data. Use the game matrix calculator to construct the heat source competition relationship matrix, calculate the game balance solution according to the optimized heat source output data to obtain the output balance state data among heat sources. Based on the output balance state data and the heat source priority sorting table, generate the final output allocation scheme, verify the feasibility of the scheme through the operation constraint verifier, and output the heat source output scheduling sequence.

[0011] Further, according to the optimized output allocation scheme, summarize the total heat source emissions. If the total emissions exceed the preset threshold, reduce the output proportion of high-emission heat sources through an iterative adjustment algorithm, and output the target output allocation matrix, including: extracting the initial heat source output value according to the optimized output allocation scheme, obtaining the pollutant concentration data per unit time from the environmental monitoring device, using the emission factor calculator to obtain the unit emissions of the heat source, and generating a dataset of the total heat source emissions. Analyze the dataset of the total heat source emissions using the concentration threshold comparator, read the emission limit and the output proportion reference value from the parameter library, and obtain the heat source emission exceedance data through cumulative emissions calculation. Sort the heat sources according to the heat source emission exceedance data, set the emissions in descending order to obtain the heat source emission ranking table, and generate the adjustment coefficient for high-emission heat sources using the output adjustment calculator. Adopt deep reinforcement learning to reduce the output of high-emission heat sources, limit the reduction step size through the response rate constraint, and calculate the output compensation value of other heat sources based on the upper limit of the heat source capacity. Optimize the output adjustment scheme using the iterative calculator, set the single-iteration adjustment step size and the maximum number of iterations, and generate a heat source output adjustment sequence. Establish the target allocation matrix according to the heat source output adjustment sequence, verify the rationality of the matrix through the operating parameter constraint checker, and output the final target output allocation scheme.

[0012] Further, identify the grid interaction attributes corresponding to the output allocation matrix, and identify the heat source output volatility. If the volatility is lower than the preset threshold, transmit the output allocation matrix to the heat source control system to determine the heat source output scheduling instruction, including: reading the heat source output data from the target output allocation matrix, obtaining the grid frequency data and time-of-use electricity price data using the grid monitoring device, generating the grid interaction index through the interaction parameter calculator, and establishing the grid interaction feature set. Perform time-frequency analysis on the heat source output data using the wavelet decomposition algorithm, extract the frequency fluctuation feature from the grid interaction feature set, and calculate the heat source output volatility according to the load change curve. Conduct a threshold check on the heat source output volatility. If the volatility is lower than the preset volatility threshold, read the control reference parameters from the preset parameter library to generate the scheduling control parameter set. Use a recursive neural network to dynamically predict the scheduling control parameter set, calculate the heat source response duration in combination with the heat source response characteristic curve, and construct the scheduling control sequence. Generate the heat source scheduling instruction according to the scheduling control sequence, verify the data transmission channel using the communication status detector, and send the scheduling instruction to the heat source control device. Obtain the execution status data through the heat source actuator feedback interface, and monitor the execution status of the heat source scheduling instruction in real time to confirm the instruction execution completion status.

[0013] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: The present invention discloses an optimization method for a heating system, aiming at the heat source scheduling problem under sudden temperature drops and equipment failures. The method first obtains the real-time operating state of the system, environmental constraint conditions and characteristics of emergencies, and determines the heat source priority; then optimizes the heat source ranking according to the cost function and environmental constraints, and initially allocates the heat source output through the dynamic programming algorithm; then combines the interactive attributes of the power grid and user demand prediction to construct a game model to optimize the output allocation; finally, considering the emission limit, iteratively adjusts the output of high-emission heat sources to obtain the final scheduling plan. Through multi-objective optimization and game analysis, the present invention realizes the comprehensive scheduling of cost minimization, environmental friendliness and power grid friendliness while ensuring the heating demand, and improves the economy, environmental protection and flexibility of the heating system BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flowchart of an optimization method for a heating system of the present invention

[0015] Figure 2 It is a schematic diagram of an optimization method for a heating system of the present invention DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative work shall fall within the protection scope of the present invention

[0017] Such as Figure 1-2 , an optimization method for a heating system in this embodiment may specifically include: S101, obtaining the real-time heat source operating state, power grid load, user demand and environmental constraint conditions of the heating system, extracting the characteristic information of sudden temperature drops and equipment failure information therefrom, and determining the initial values of the heat source priority and response speed

[0018] In the embodiment of the present application, the heating system is configured with an optimization function, which can be integrated into the system control software or run independently in a dedicated application program. The specific implementation form can be flexibly set according to the actual scenario. The startup method of the optimization function is not strictly limited. The user can trigger it through manual operation or remotely activate it by an external device sending an instruction. For example, when the optimization function is embedded in the system control software, the user can select the startup option in the interface; when implemented by an application program, it can be quickly enabled by clicking a preset icon

[0019] S1011. Collect the real-time heat source operation status data set, the load data set of the power grid monitoring unit, the total heat consumption data set of the intelligent meters at the user end, and the pollutant concentration and carbon emission concentration data set of the environmental monitoring device through the heating system monitoring device, and update the heat source operation status data in real time.

[0020] In the embodiment of the present application, the heating system monitoring device is equipped with multiple temperature sensors, with a temperature measurement range of 0 to 150 degrees Celsius, a sampling frequency of 1 time per second, and a data storage period of up to 30 days. The power grid monitoring unit obtains the load data through a power collector, with a sampling frequency of 1 time per minute. The intelligent meters at the user end are deployed according to the heat meter standard, and the accuracy of recording the heat consumption reaches 0.1 kWh. The environmental monitoring device is arranged around the heat source, and the monitoring accuracy of carbon dioxide concentration is 0.1 mg per cubic meter. These data provide basic support for the analysis.

[0021] S1012. Extract the temperature monitoring data from the heating system monitoring device, calculate the temperature change rate using the central difference method based on the temperature values at multiple monitoring points, judge the sudden cooling event through a preset threshold, generate the cooling amplitude data set and the cooling rate data set, and at the same time standardize the heat source operation status data. If an abnormal fluctuation is detected, use the pre-trained deep neural network to identify the fault type and combine the fault diagnosis rule base to quantify the fault severity, and obtain the fault severity level data.

[0022] In the embodiment of the present application, the temperature change rate is calculated by dividing the temperature difference between adjacent moments by the time interval, and the set threshold is 5 degrees Celsius per hour. If it exceeds, it is determined as a sudden cooling event. The heat source operation status data includes multi-dimensional parameters such as pressure, flow rate, and temperature, and the dimension is unified after standardization. If the data shows an abnormal fluctuation, the pre-trained neural network can identify types such as valve failure, pump station failure, or pipeline network leakage. The fault severity is divided into 1 to 10 levels, and the higher the value, the greater the impact.

[0023] S1013. Build a heat source response model based on the cooling amplitude data set, the cooling rate data set, and the fault severity level data, dynamically adjust the heat source response speed through the response speed compensation algorithm, and generate the heat source priority data set and the initial response speed data set.

[0024] In the embodiment of the present application, the heat source response model comprehensively analyzes the cooling characteristics and fault conditions, and calculates the supply priority of each heat source. For example, in a sudden cooling event in a certain area, the cooling amplitude is 8 degrees Celsius, the rate is 6 degrees Celsius per hour, and the pump station fault level reaches 7. The model outputs the highest priority and the response speed is 4 degrees Celsius per hour. The response speed compensation algorithm adjusts the coefficient between 0.8 and 1.2 according to the actual working conditions to ensure the stable operation of the system and generate the initial data set, laying a foundation for optimization.

[0025] In the embodiments of the present application, through the above steps, key information can be quickly extracted and heat source parameters can be initialized in complex scenarios, providing a reliable basis for multi-heat source collaborative scheduling.

[0026] S102. Calculate the heating interruption ratio according to the sudden cooling characteristic information and equipment failure data. When the interruption ratio exceeds the preset threshold, determine the upper limit of the heat source output and adjust the heat source priority sequence based on historical data and the current low-load state.

[0027] In the embodiments of the present application, the heating system needs to cope with the heating interruption risks brought by sudden cooling and equipment failures. For this purpose, the system will monitor the environmental temperature and equipment status in real time and perform intelligent scheduling based on this information. Specifically, relevant data can be collected and analyzed through a heating monitoring device, and the specific method is set by technicians according to the actual scenario.

[0028] S1021. Construct a heating interruption weight coefficient matrix according to the sudden cooling amplitude data and fault duration data. Generate an initial value of the interruption ratio for the heating area through matrix operations. At the same time, use a time-series cumulative calculator to record the interruption ratio data at each moment and obtain the interruption ratio threshold from a preset parameter library.

[0029] In the embodiments of the present application, the heating interruption weight coefficient matrix is used to quantify the impact of sudden cooling and equipment failures on heating capacity. The matrix is constructed based on two dimensions: cooling amplitude and fault duration. For example, for every 5-degree Celsius increase in the cooling amplitude, the weight increases by 0.2, and for every 1-hour extension of the fault duration, the weight increases by 0.1. Suppose the cooling amplitude in a certain area is 15 degrees Celsius and the fault duration is 4 hours. Then the initial interruption ratio calculated by the matrix is 0.6 times 15 divided by the maximum amplitude range of 25 plus 0.4 times 4 divided by the maximum duration of 10, and the result is approximately 0.52. The time-series cumulative calculator records the change in the interruption ratio in minutes, and the preset threshold is usually set to 0.5 and can be adjusted according to system requirements. This process ensures the dynamic tracking and accurate assessment of the interruption ratio.

[0030] S1022. Obtain the load fluctuation data from the heating load acquisition device. Generate a smoothed load curve by setting a 24-hour sliding time window and using the exponentially weighted moving average method. Divide the load curve period based on the density clustering algorithm and extract the data of the load valley period.

[0031] In the embodiment of the present application, the load fluctuation data is acquired by the heating load acquisition device at a sampling interval of 5 minutes. To reduce the influence of data noise, the exponential weighted average method is used for smoothing, and the smoothing coefficient is set to 0.3 to assign higher weights to recent data. The smoothed load curve is divided into time periods through the density clustering algorithm, which identifies troughs based on the aggregation characteristics of load values. For example, the load from 2 am to 6 am drops to 65% of the daily average value and is determined as the trough period. This step provides a reliable basis for load trend analysis.

[0032] S1023, normalize the load records in the historical database for the past 30 days to generate standardized load data, calculate the load change trend curve through the support vector regression algorithm, determine the heat source operation range by combining the current load value and the load trough period data, and construct a heat source output limit matrix based on this range.

[0033] In the embodiment of the present application, the historical load data is mapped to the range of 0 to 1 after normalization to eliminate the dimension difference. The support vector regression uses a radial basis kernel function, and the kernel parameter is set to 0.1 to generate the load change trend curve through training. For example, the current load value is 80 MW, and the average load during the trough period is 52 MW. The predicted operation range of the trend curve is 45 to 85 MW. The heat source output limit matrix sets the output range of each heat source according to this range, and at the same time considers constraints such as the rated capacity and operation efficiency to provide a basis for output calculation.

[0034] S1024, if the initial value of the heating area interruption ratio exceeds the preset threshold, calculate the adjustment rate correction coefficient according to the heat source operation range data and the heat source response time curve, jointly determine the upper limit value of the heat source output through the heat source output limit matrix and the correction coefficient, and generate an adjusted priority sequence in combination with the heat source priority reference data.

[0035] In the embodiment of the present application, when the interruption ratio, such as 0.52, exceeds the threshold of 0.5, the system enters the output adjustment mode. The heat source response time curve reflects the influence of the load change rate on the adjustment rate. For example, when the change rate exceeds 20% per hour, the correction coefficient drops to 0.8. Combining the output limit matrix, the upper limit of the heat source output is calculated. For example, the upper limit of a certain system is 75 MW. The priority reference data includes the rated capacity, efficiency, and start-stop characteristics of the heat sources. Assuming that the capacities of three heat sources are 50 MW, 30 MW, and 20 MW respectively, after adjustment, the 50 MW unit has the highest priority, and the output upper limit is set to 45 MW. The 30 MW unit follows, with an output upper limit of 20 MW, and the 20 MW unit is used as a standby. This adjustment ensures the flexibility of the system to meet the demand while having adjustment flexibility.

[0036] In this embodiment of the present application, the above steps enable precise calculation of the interruption ratio and dynamic optimization of heat source output. In particular, in scenarios where sudden temperature drops and faults coexist, the system can quickly respond and adjust priorities. For example, maintaining the total output of a certain area at 65 MW not only meets the current load but also reserves room for future changes. The specific parameters and algorithm implementation can be further optimized by technical personnel based on actual operating conditions and are not detailed here.

[0037] S103, obtaining the required parameters of the cost function corresponding to the adjusted heat source priority sequence, determining the cost weight of each heat source through a linear weighting method and outputting the optimized cost ranking result.

[0038] In this embodiment, after adjusting the heat source priority, the scheduling order needs to be further optimized based on cost factors. The acquisition and calculation of cost parameters aims to balance fuel consumption, operational efficiency, and environmental requirements, ensuring both system economics and sustainability. The specific implementation method can be flexibly adjusted based on the actual heat source type and operating environment.

[0039] S1031, extract the fuel unit price data corresponding to the adjusted priority sequence from the heat source database, collect equipment operation efficiency data through the heat source operation monitoring device and obtain emission factor data from the pollutant online monitor, and construct a cost parameter data set containing multiple heat source types.

[0040] In an embodiment of the present application, the heat source database stores basic information of various heat sources. For example, a system includes coal-fired boilers, gas boilers, and electric boilers, with fuel unit prices of 800 yuan per ton, 3.2 yuan per cubic meter, and 0.65 yuan per kilowatt-hour, respectively. The operation monitoring device records the equipment efficiency in real time, which is 82%, 92%, and 98%, respectively. The pollutant monitor measures the emission factor of 2.62 kilograms per ton, 1.8 kilograms per cubic meter, and 0 kilograms per kilowatt-hour. These data are collected and unified through an integrated interface to form a basic data set for cost analysis.

[0041] S1032, based on the heat source type and equipment operating status in the cost parameter data set, uses the standardized unit conversion method to calculate the unit time fuel consumption, cumulative operating time and standard emission concentration data, and uses the multiple linear regression method to generate the initial cost weight coefficient, and optimizes the weight through residual analysis to generate the standardized cost weight coefficient.

[0042] In the embodiments of the present application, the fuel consumption per unit time is related to the heat source load. For example, a coal-fired boiler consumes 4.2 tons per hour at 50% load, a gas-fired boiler consumes 380 cubic meters per hour, and an electric boiler consumes 2,800 kWh per hour. The cumulative operating hours are extracted from the equipment logs and are 2,160 hours, 1,440 hours, and 720 hours respectively. The emissions are converted to sulfur dioxide equivalent through standardization, and the results are 145 mg / m³, 85 mg / m³, and 0 mg / m³. Multiple linear regression uses the least squares method to calculate the initial weights, setting the fuel cost weight to 0.5, the operating cost to 0.3, and the environmental protection cost to 0.2. Residual analysis found that the fuel cost deviation was about 15%, and after optimization, the weights were adjusted to 0.45, 0.35, and 0.2 to ensure that the calculation results are closer to the actual operating conditions.

[0043] S1033, use the standardized cost weight coefficient to perform weighted combination on the fuel consumption, operating duration, and emission concentration data to generate a comprehensive cost function value, establish a normalization processing matrix through linear weighted calculation in combination with the actual heat source output data, and generate a heat source operation priority lookup table based on this to update the scheduling order.

[0044] In the embodiments of the present application, when calculating the comprehensive cost by weighted calculation, the fuel cost of the coal-fired boiler is 3,360 yuan per hour, the operating cost is 420 yuan per hour, the environmental protection cost is 580 yuan per hour, and the total cost is 4,360 yuan per hour; the total cost of the gas-fired boiler is 2,860 yuan per hour, and the electric boiler is 2,420 yuan per hour. Normalization maps the cost to the interval from 0 to 1, and the results are 1, 0.66, and 0.56. Combining the actual output of 20,000 kW, 15,000 kW, and 10,000 kW, a priority lookup table is generated after linear weighting. The electric boiler has the highest priority, the gas-fired boiler follows, and the coal-fired boiler has the lowest priority. The new scheduling order is that the electric boiler undertakes the base load, the gas-fired boiler performs peak shaving, and the coal-fired boiler is in standby. This method optimizes the resource allocation and improves the economic benefits.

[0045] In the embodiments of the present application, accurate acquisition and optimization sorting of cost parameters are achieved through the above steps. For example, in a scenario where the heating load is 45,000 kW, the high efficiency and zero-emission characteristics of the electric boiler enable it to operate preferentially, while the coal-fired boiler is placed in standby due to high costs and emissions. This process not only reduces the operating costs but also reduces the environmental load. The specific weights and calculation details can be further adjusted by technicians according to actual needs and will not be elaborated here.

[0046] In practical applications, if a certain area needs to quickly respond to the heating demand, the output can be adjusted in real time according to the priority lookup table. For example, the electric boiler operates at full load during low electricity price periods, and the gas-fired boiler flexibly supplements the peak demand to ensure heating stability while minimizing costs.

[0047] S104. Evaluate the power grid load status according to cost optimization sorting and environmental constraint conditions. If the power grid load is lower than the preset threshold, calculate the heat source output distribution plan through the dynamic programming algorithm and generate a preliminary output matrix.

[0048] In the embodiment of the present application, the heat source output distribution needs to comprehensively consider economy and environmental protection. Especially in the scenario where the power grid load fluctuates greatly, through real-time monitoring and intelligent algorithm optimization scheduling, ensure stable heating and meet environmental requirements. The implementation process can be flexibly adjusted according to the actual system configuration.

[0049] S1041. Obtain the real-time power grid load value and load fluctuation data from the power grid monitoring device, extract the sorting number in combination with the cost optimization sorting result, and at the same time collect the pollutant concentration data through the environmental monitoring device to jointly construct a load evaluation data set.

[0050] In the embodiment of the present application, the power grid monitoring device collects load data at intervals of 5 minutes, records the real-time value and fluctuation conditions. For example, the load in a certain area drops to 60% of the daily average at night, while it can reach 145% during peak hours. The environmental monitoring device records the pollutant concentration once an hour, covering key indicators such as sulfur dioxide, nitrogen oxides, and soot. These data are integrated through the interface into a load evaluation data set, providing a basis for subsequent analysis.

[0051] S1042. Perform normalization processing on the load evaluation data set through a data filter to map the data to the interval from 0 to 1, use support vector regression to predict the load fluctuation trend, calculate the environmental constraint index according to the pollutant concentration, and at the same time use a recursive neural network to match the heat resource capacity and heating load to determine the adjustment speed and distribution weight.

[0052] In the embodiment of the present application, the normalization processing eliminates the dimension difference, facilitating cross-period comparison. Support vector regression uses a radial basis kernel function for modeling and predicts the load trend in the next 4 hours. For example, it predicts that the load will continue to be low at night. The environmental constraint index is calculated based on pollutant limits, such as a sulfur dioxide limit of 35 mg / m³, a nitrogen oxide limit of 50 mg / m³, and a soot limit of 10 mg / m³. The recursive neural network processes data through long short-term memory units, matches the capacities of 3 units (20,000 kW, 15,000 kW, 10,000 kW) with the load, and obtains adjustment speeds of 2000 kW per hour, 1500 kW per hour, and 1000 kW respectively. The distribution weight is dynamically adjusted according to the load interval, and the weight of the large-capacity unit is higher at high loads.

[0053] S1043. If the power grid load is lower than the preset threshold, use the dynamic programming algorithm to optimize the heat source output according to the adjustment speed and distribution weight, and generate a preliminary output matrix in combination with the environmental constraint index to ensure the balance between heating demand and environmental protection goals.

[0054] In this application scenario, assume that the load threshold is 80,000 kW and the current load is 60,000 kW, triggering the optimization process. The dynamic programming algorithm aims to minimize costs and emissions. Combining environmental data (sulfur dioxide at 28 mg per cubic meter, nitrogen oxides at 42 mg per cubic meter, and soot at 8 mg per cubic meter), it allocates output for a load of 45,000 kW: the 20,000 kW unit outputs 18,000 kW, the 15,000 kW unit outputs 12,000 kW, and the 10,000 kW unit outputs 5,000 kW. This solution avoids frequent startups and shutdowns and meets environmental protection requirements.

[0055] In actual operation, it can be flexibly adjusted according to the load peak and valley characteristics. For example, during low load periods, electric boilers are preferentially started, and during peak load periods, the output of gas boilers is increased. This method not only optimizes resource utilization but also reduces operating costs and pollutant emissions. Specific parameters can be further adjusted according to actual needs.

[0056] S105. Obtain the grid interaction attributes, combine the preliminary output matrix with the user-side demand response data, extract the peak and valley fluctuation trends of heat consumption through time series analysis, and predict short-term user demands, and output the demand response adjustment coefficient to optimize the scheduling.

[0057] In the embodiments of this application, the combination of grid interaction attributes and user demand response aims to enhance the adaptability of the heating system to external condition changes. By analyzing the grid state and user behavior, the system can dynamically adjust the output allocation, not only meeting the heating demand but also making full use of the electricity price advantage and grid stability. The specific implementation can be adjusted according to the actual scenario requirements.

[0058] S1051. Extract the time-of-use electricity price data and grid frequency data from the grid management center, combine them with the preliminary output matrix to calculate the grid interaction index. At the same time, collect the real-time heat consumption and user room temperature data through intelligent heat meters to construct a user load dataset including the time dimension and user behavior.

[0059] In the embodiments of this application, the grid management center provides time-of-use electricity price data. One day is divided into peak periods at 1.2 yuan per kWh, normal periods at 0.8 yuan per kWh, and valley periods at 0.4 yuan per kWh. The grid frequency data reflects stability, with a normal value of 50 Hz and a fluctuation range of 49.8 to 50.2 Hz. The preliminary output matrix shows that the output ratio of the electric boiler is high during valley periods, and the calculated grid interaction index is 0.82. The intelligent heat meter records the heat consumption and room temperature data every 15 minutes. For example, in a certain community, the heat consumption from 6 to 9 am during the morning peak reaches 160% of the daily average value, and drops to 50% during the night low period, and the room temperature is maintained at 18 to 24 degrees Celsius. These data lay the foundation for subsequent analysis.

[0060] S1052. Reconstruct the time series of the user load dataset and use a recurrent neural network to extract the heat consumption fluctuation features. Identify the peak-valley load intervals through time period division markers and generate a fluctuation curve. At the same time, standardize the room temperature data and the time-of-use electricity price data, calculate the user response eigenvalue in combination with the fluctuation curve, and establish a response dataset.

[0061] In the embodiment of the present application, the recurrent neural network extracts the fluctuation features by analyzing the historical heat consumption data, such as identifying the double-peak pattern in the early and late hours of weekdays. The time period division markers divide a day into peak, flat, and valley intervals, and the generated fluctuation curve shows that the peak hours are from 7:00 to 9:00 in the morning and from 19:00 to 22:00 in the evening. After standardizing the room temperature and electricity price data, they are mapped to the interval from 0 to 1. When calculating the response eigenvalue, it is found that the user sensitivity increases when the room temperature is lower than 20 degrees Celsius, and the heat consumption decreases during the electricity price peak. This analysis reveals the dual response characteristics of users to temperature and price, providing a basis for demand prediction.

[0062] S1053. Use a long short-term memory network to perform time series prediction on the user response dataset and set a sliding time window to obtain short-term demand prediction data. Construct a response feature matrix based on the peak-valley load intervals and the prediction data, extract the user response pattern through matrix decomposition and calculate the benchmark value, and finally generate a demand response adjustment coefficient in combination with the grid interaction index.

[0063] In the embodiment of the present application, the long short-term memory network adopts a three-layer structure, inputs the data of the recent 7 days, has 128 neurons in the hidden layer, and predicts the demand for the next 24 hours. The results show that the demand during the morning peak is strongly periodic and sensitive to the room temperature. The response feature matrix contains three dimensions: time period, temperature, and price. After decomposition, the benchmark values of the morning peak pattern are extracted as 0.85, 0.72 for the evening peak, and 0.45 for the valley period. Combining the grid interaction index of 0.82, a piecewise adjustment function is constructed: when the grid frequency is lower than 49.9 Hz, the adjustment coefficient increases to cut the load; when the electricity price is low and the frequency is normal during the valley period, the coefficient decreases to encourage heat consumption. For example, on a certain day, the coefficient during the morning peak is 1.15, 1.08 for the evening peak, and 0.92 for the valley period, ensuring the coordinated optimization of heat supply and the grid.

[0064] In practical applications, this method can be dynamically adjusted according to user habits and grid conditions. For example, during low electricity price periods, it encourages electric boilers to generate more power, and during peak periods, it guides users to cut peaks and fill valleys, which not only improves the system efficiency but also reduces the operating cost. The specific parameters can be further optimized according to the actual operating environment.

[0065] S106. Combine the preliminary output matrix and the demand response adjustment coefficient to construct a game model, extract game parameters from the heat source priority, cost function, and user demand, and generate an optimized output allocation plan by calculating the output game results among the heat sources.

[0066] In this application scenario, the game model aims to coordinate the output distribution among multiple heat sources, balance costs, efficiency, and user demands. By introducing intelligent algorithms, it can dynamically adapt to complex working conditions to ensure the stability and economy of heat supply. The specific implementation method can be adjusted according to the heat source configuration and operation requirements.

[0067] S1061, Extract the heat source output reference data from the preliminary output matrix and obtain the weight information from the priority sequence. Combine the cost function value and the user demand quantity to construct the initial game revenue matrix. At the same time, set the game constraint conditions based on the demand response adjustment coefficient to reflect the influence on the user side.

[0068] In the embodiment of this application, the preliminary output matrix records the reference output of each heat source. For example, the electric boiler is 10,000 kW, the gas boiler is 15,000 kW, and the coal-fired boiler is 20,000 kW. The priority weights reflect the operation sequence, which are 0.5, 0.3, and 0.2 respectively. Combine the cost function (the hourly operation costs are 280 yuan, 420 yuan, and 580 yuan respectively) and the user demand of 45,000 kW to construct the initial matrix. The demand response adjustment coefficient such as 1.15 is introduced into the constraint to reflect the user's response to price or temperature changes and ensure that the game result is close to the actual demand.

[0069] S1062, Obtain the upper limit value of the heat supply capacity and the response rate limit value from the heat source operation database and generate a set of game parameters in combination with the constraint conditions. Normalize the set of parameters to obtain the standardized game parameters. Subsequently, use deep reinforcement learning to train the parameters and calculate the competitive strategy revenue of each heat source through the reward function.

[0070] In this application scenario, the database provides operation parameters: the upper limit of the electric boiler is 12,000 kW, and the response rate is 10,000 kW per hour; the upper limit of the gas boiler is 20,000 kW, and the response rate is 8,000 kW per hour; the upper limit of the coal-fired boiler is 25,000 kW, and the response rate is 5,000 kW per hour. After normalizing the parameters to the range of 0 to 1, deep reinforcement learning is trained through the reward function. The function comprehensively considers costs, emissions, and regulation performance, and obtains a positive reward when the output meets the demand, and is punished when it exceeds the limit, generating revenue data to reflect the advantages and disadvantages of each heat source's strategy.

[0071] S1063, Construct the heat source competition game matrix based on the competitive strategy revenue data and calculate the equilibrium solution through the Nash equilibrium solver. Dynamically adjust the equilibrium solution using an adaptive neural network and verify it based on the capacity upper limit and rate limit. Optimize the actual output of each heat resource through multi-objective programming.

[0072] In the embodiment of the present application, the game matrix is constructed based on the revenue data, and the Nash equilibrium calculates the initial plan: 12,000 kilowatts for the electric boiler, 20,000 kilowatts for the gas boiler, and 13,000 kilowatts for the coal-fired boiler. The adaptive neural network adjusts according to the real-time data. The output of the electric boiler is reduced to 11,000 kilowatts due to the rate limit, the gas boiler is adjusted to 18,000 kilowatts, and the coal-fired boiler is increased to 16,000 kilowatts. The multi-objective programming optimizes the total output to 45,000 kilowatts, taking into account both cost and environmental protection to ensure the feasibility of the plan.

[0073] In another implementation path, the heat resource priority and cost data are extracted from the database and the initial weights are calculated. The output distribution is predicted through a recursive neural network and threshold verification is performed. If the threshold is exceeded, the weights are reallocated, and a scheduling sequence is generated through game balance solution. For example, the weight of the electric boiler is adjusted to 0.35, the gas boiler is increased to 0.4, and the coal-fired boiler is 0.25. The predicted output forms a sequence after verification: 11,000 kilowatts for the electric boiler, 18,000 kilowatts for the gas boiler, and 16,000 kilowatts for the coal-fired boiler, meeting the requirements and conforming to the constraints.

[0074] In actual operation, this method improves the scheduling flexibility through the combination of game and prediction. For example, the electric boiler is preferentially started during high demand periods, and the output of the gas boiler is increased during low-cost periods, which not only reduces the operating costs but also ensures the heating stability. The specific parameters can be further adjusted according to the working conditions.

[0075] S107. Calculate the total heat source emissions according to the optimized output distribution plan. If it exceeds the preset threshold, reduce the output proportion of the high-emission heat source through the iterative adjustment algorithm and output the target output distribution matrix.

[0076] In this application scenario, emission control is a key link in the optimization of the heating system. By monitoring the pollutant emissions in real time and dynamically adjusting the output, the environmental protection goals can be achieved while meeting the heating demand. The specific implementation can be flexibly adjusted according to the heat source type and emission characteristics.

[0077] S1071. Extract the initial output data of the heat source according to the optimized output distribution plan, obtain the pollutant concentration data per unit time from the environmental monitoring device, and generate the total heat source emissions dataset using the emission factor calculator. Subsequently, analyze the emission exceedance data through the concentration threshold comparator and the preset emission limit.

[0078] In the embodiment of the present application, the initial output plan is 16,000 kilowatts for the coal-fired boiler, 18,000 kilowatts for the gas-fired boiler, and 11,000 kilowatts for the electric boiler. The environmental monitoring device collects pollutant concentrations once an hour, including sulfur dioxide, nitrogen oxides, and soot. The emission factor calculator generates a total emission dataset based on the unit emissions (2.62 kilograms per kilowatt-hour for the coal-fired boiler, 0.82 kilograms per kilowatt-hour for the gas-fired boiler, and 0 kilograms per kilowatt-hour for the electric boiler). The concentration threshold comparator analyzes with reference to the limit values in the parameter library, such as 35 milligrams per cubic meter for sulfur dioxide, 50 milligrams per cubic meter for nitrogen oxides, and 10 milligrams per cubic meter for soot, and shows that the hourly emissions of the coal-fired boiler are 4,192 kilograms, and the total emissions exceed the standard.

[0079] S1072, sort the emission-exceeding data in descending order to generate a heat source emission level table, determine the adjustment coefficient of the high-emission heat source through the output adjustment calculator, then perform output reduction using deep reinforcement learning, and calculate the output compensation value of other heat sources in combination with the response rate constraint.

[0080] In this application scenario, the heat source emission level table is sorted as coal-fired boiler, gas-fired boiler, and electric boiler, and the coal-fired boiler is preferentially adjusted due to high emissions. The output adjustment calculator sets the coefficient of the coal-fired boiler to 0.8, and 20% of the output needs to be reduced. Deep reinforcement learning optimizes the reduction strategy through the reward function, and the reward function encourages low emissions and high efficiency. The single reduction is limited by the response rate, for example, 5,000 kilowatts per hour for the coal-fired boiler. The compensation value is calculated based on the upper limit of 20,000 kilowatts for the gas-fired boiler and the upper limit of 12,000 kilowatts for the electric boiler to ensure the stability of the total output.

[0081] In subsequent optimization, the iteration calculator adjusts the output in steps of 2,000 kilowatts and a maximum of 10 iterations. The final plan is 12,000 kilowatts for the coal-fired boiler, 20,000 kilowatts for the gas-fired boiler, and 12,000 kilowatts for the electric boiler. The target matrix is verified to confirm that the total heat supply is 44,000 kilowatts, the emissions are reduced by 28%, meeting the environmental protection requirements and the output change rate conforming to the constraints. This method significantly improves the environmental protection performance of the system through intelligent adjustment while ensuring the reliability of heat supply. The specific limit values and step sizes can be further optimized according to actual needs.

[0082] S108, identify the grid interaction attributes corresponding to the target output allocation matrix and calculate the heat source output volatility. If the volatility is lower than the preset threshold, transmit the matrix to the heat source control system and generate a scheduling instruction to achieve precise control.

[0083] In this application scenario, the analysis of grid interaction attributes and output volatility aims to ensure the stability and adaptability of the scheduling plan. Through real-time monitoring and algorithm analysis, the system can efficiently execute output allocation in a complex grid environment. The specific implementation method can be adjusted according to actual equipment and grid conditions.

[0084] S1081. Extract the heat source output data from the target output distribution matrix, combine the frequency and time-of-use electricity price data obtained by the power grid monitoring device, generate the power grid interaction index through the interaction parameter calculator, and establish a feature set. Subsequently, use the wavelet decomposition algorithm for time-frequency analysis to calculate the output volatility.

[0085] In the embodiment of the present application, the target output distribution matrix records the output data of the electric boiler at 12,000 kW, the gas boiler at 20,000 kW, and the coal-fired boiler at 16,000 kW. The power grid monitoring device provides frequency data in the range of 49.8 to 50.2 Hz, and the time-of-use electricity prices are 1.2 yuan per kWh during the peak period, 0.8 yuan per kWh during the normal period, and 0.4 yuan per kWh during the valley period. The interaction parameter calculator comprehensively evaluates to obtain an index of 0.85, indicating that the power grid state is stable. The wavelet decomposition algorithm decomposes the 24-hour output data into high-frequency short-term fluctuations and low-frequency trend components, combines the frequency fluctuation of 0.2 Hz and the load change curve, and calculates the output volatility to be 8%, mainly due to the rapid adjustment of the electric boiler.

[0086] S1082. If the heat source output volatility is lower than the preset threshold, extract the control reference parameters from the parameter library and generate a set of dispatching control parameters. Predict the dynamic changes of the parameters through the recurrent neural network and calculate the response duration in combination with the response characteristic curve, and generate and issue a dispatching instruction.

[0087] In this application scenario, the preset fluctuation threshold is 10%, and the current volatility of 8% meets the requirements. The parameter library provides control references, such as the adjustment rate limit of the electric boiler being 10,000 kW per hour, the gas boiler being 8,000 kW per hour, and the coal-fired boiler being 5,000 kW per hour. The recurrent neural network predicts the future 4-hour load trend based on 72-hour historical data, showing a stable change. The response characteristic curve calculates that the response duration of the electric boiler is 5 minutes, the gas boiler is 15 minutes, and the coal-fired boiler is 30 minutes. The generated control sequence is sorted according to the response speed and issued through the communication status detector. The detection result shows that the delay is less than 100 milliseconds and the error rate is less than 0.01%.

[0088] In actual execution, the feedback data of the heat source actuator shows that the electric boiler quickly reaches 12,000 kW, the gas boiler adjusts at a rate of 7,000 kW per hour, and the coal-fired boiler stabilizes at 16,000 kW. All parameters meet the limits. This method ensures the efficient execution of the dispatching instruction and the system stability, and is applicable to various operating scenarios. The specific threshold and prediction period can be optimized according to requirements.

[0089] The above are only the preferred embodiments of one or more embodiments of this specification, and are not intended to limit one or more embodiments of this specification. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of one or more embodiments of this specification shall be included within the scope protected by one or more embodiments of this specification.

Claims

1. An optimization method for a heating system, characterized in that The method includes: Obtain the real-time heat source operation status, grid load, user demand, and environmental constraint conditions of the heating system, extract the characteristic information of sudden temperature drops and equipment failure information, and determine the initial values of heat source priorities and response speeds; According to the characteristic information of sudden temperature drops and equipment failure data, calculate the interruption ratio. If the interruption ratio exceeds the preset threshold, calculate the upper limit of heat source output through historical data and the current low-load state of the grid, and adjust the heat source priority sequence; Obtain the parameters required for the cost function corresponding to the adjusted heat source priority sequence, determine the cost weights of different heat sources after adjustment through the linear weighting method, and output the cost-optimized sorting; Identify the grid load according to the cost-optimized sorting and environmental constraint conditions. If the grid load is lower than the preset threshold, calculate the initial heat source output allocation plan through the dynamic programming algorithm to obtain the preliminary output matrix; Obtain the grid interaction attributes, combine the preliminary output matrix with the user-side demand response data, extract the fluctuation trend of peak-valley characteristics from the real-time heat consumption changes, determine the short-term prediction value of user demand through time series analysis, and output the demand response adjustment coefficient; Combine the preliminary output matrix with the demand response adjustment coefficient, construct a game model, extract game parameters from the heat source priority, cost function, and user demand, calculate the output game results between heat sources, and obtain the optimized output allocation plan; According to the optimized output allocation plan, summarize the total heat source emissions. If the total emissions exceed the preset threshold, reduce the output proportion of high-emission heat sources through the iterative adjustment algorithm, and output the target output allocation matrix; Identify the grid interaction attributes corresponding to the output allocation matrix, identify the heat source output volatility. If the volatility is lower than the preset threshold, transmit the output allocation matrix to the heat source control system to determine the heat source output scheduling instruction.

2. The method according to claim 1, wherein The obtaining of the real-time heat source operation status, grid load, user demand, and environmental constraint conditions of the heating system, extracting the characteristic information of sudden temperature drops and equipment failure information, and determining the initial values of heat source priorities and response speeds includes: Collect the real-time heat source operation status data set and temperature monitoring data from the heating system monitoring device, calculate the temperature change rate using the central difference method according to the temperature monitoring data, judge the sudden temperature drop event through the temperature change rate threshold, and obtain the temperature drop amplitude data set and temperature drop rate data set; Perform standardization processing on the heat source operation status data set. If there are abnormal fluctuations in the standardized data, use a pre-trained deep neural network to identify the fault type, and quantitatively evaluate the fault severity through the fault diagnosis rule base to obtain the fault severity level data; Establish a heat source response model according to the temperature drop amplitude data set, temperature drop rate data set, and fault severity level data, and dynamically adjust the heat source response speed data using the response speed compensation algorithm to obtain the heat source priority data set and the initial response speed data set.

3. The method according to claim 1, wherein The calculating of the interruption ratio according to the characteristic information of sudden temperature drops and equipment failure data, and if the interruption ratio exceeds the preset threshold, calculating the upper limit of heat source output through historical data and the current low-load state of the grid, and adjusting the heat source priority sequence includes: Establish a heat supply interruption weight coefficient matrix based on the sudden temperature drop amplitude data and the fault duration data, and calculate the initial value of the interruption ratio of the heat supply area through the heat supply interruption weight coefficient matrix; Obtain the load fluctuation data from the heat supply load acquisition device, use the exponentially weighted moving average method for the load fluctuation data to obtain a smoothed load curve, and use the density clustering algorithm to divide the smoothed load curve into time periods to obtain the load valley period data; Calculate the load change trend curve according to the load valley period data through support vector regression, and calculate the heat source operation interval data by combining the load change trend curve and the current load value data; If the initial value of the heat supply area interruption ratio exceeds the interruption ratio threshold data, then establish a heat source output limit matrix based on the heat source operation interval data, calculate the adjustment rate correction coefficient through the heat source response time curve, calculate the upper limit value of the heat source output according to the heat source output limit matrix and the adjustment rate correction coefficient, and generate a priority dynamic adjustment sequence in combination with the heat source priority benchmark.

4. The method according to claim 1, wherein Obtain the parameters required for the cost function corresponding to the adjusted heat source priority sequence, determine the cost weights of different heat sources after adjustment by the linear weighting method, and output the cost optimization sorting, including: Read the heat source priority number and fuel unit price value from the heat source database, and obtain the equipment operation efficiency data and emission factor data through the heat source operation monitoring device to obtain a cost parameter data set; According to the heat source type and equipment operation status parameters in the cost parameter data set, use the standardized unit conversion method to calculate the fuel consumption per unit time data, the cumulative operation duration data, and the standard emission concentration data; For the fuel consumption per unit time data, the standard emission concentration data, and the cumulative operation duration data, use the multiple linear regression method to calculate the initial cost weight coefficient, and obtain the standardized cost weight coefficient through regression residual analysis; Establish a normalization processing matrix according to the standardized cost weight coefficient, perform linear weighting calculation using the actual heat source output data, obtain a heat source operation priority lookup table, and update the heat source scheduling order data through the priority mapping relationship.

5. The method according to claim 1, wherein Identify the grid load according to the cost optimization sorting and environmental constraint conditions. If the grid load is lower than the preset threshold, calculate the initial heat source output allocation plan through the dynamic programming algorithm to obtain a preliminary output matrix, including: Read the grid load value and load fluctuation value from the grid monitoring device, and construct a load evaluation data set in combination with the cost optimization sorting data; Perform normalization processing on the load evaluation data set through a data filter to obtain normalized data, and use support vector regression to predict the fluctuation trend of the normalized data to obtain an environmental constraint index; Use a recurrent neural network to perform matching calculations on the heat source capacity value and the heat supply load quantity to obtain a matching result, and obtain the adjustment speed value and the allocation weight value from the heat source parameter library according to the matching result; If the grid load value is lower than the load threshold, use the dynamic planner to optimize the calculation of the heat source output according to the adjustment speed value and the allocation weight value, and generate a preliminary output matrix according to the environmental constraint index.

6. The method according to claim 1, characterized in that Obtaining the grid interaction attributes, combining the preliminary output matrix with the user-side demand response data, extracting the fluctuation trend of peak-valley characteristics from the real-time heat consumption changes, determining the short-term prediction value of user demand through time series analysis, and outputting the demand response adjustment coefficient, including: Calculating the grid interaction index according to the time-of-use electricity price data and grid frequency data of the grid management center, and collecting the real-time heat consumption data and user room temperature data through intelligent heat meters to obtain the user load data set; Extracting the heat consumption fluctuation characteristics from the user load data set by using a recurrent neural network, and judging the peak-valley load interval according to the time period division mark to obtain the heat consumption fluctuation curve; Performing data standardization processing according to the user room temperature data and time-of-use electricity price data, calculating the user response characteristic value by using the heat consumption fluctuation curve, establishing a response characteristic matrix, and extracting the user response mode by using the matrix decomposition method to obtain the user response reference value; Based on the user response reference value and the grid interaction index, establishing a response adjustment function, and calculating the demand response adjustment coefficient through the function.

7. The method according to claim 1, wherein Combining the preliminary output matrix with the demand response adjustment coefficient, constructing a game model, extracting game parameters from the heat source priority, cost function and user demand, calculating the output game result between heat sources, and obtaining the optimized output distribution scheme, including: Constructing an initial game revenue matrix according to the user demand quantity and cost function value, and the initial game revenue matrix is generated by the heat source output matrix and the demand response adjustment coefficient; Obtaining the upper limit value of the heating capacity and the limit value of the response rate from the heat source operation database, and performing normalization processing on the initial game revenue matrix to obtain the standardized game parameters; Training the standardized game parameters by using deep reinforcement learning, and calculating the heat source competition strategy revenue data through the reward function; Performing dynamic adjustment on the heat source competition strategy revenue data by using an adaptive neural network. If the dynamic adjustment result meets the upper limit value of the heating capacity and the limit value of the response rate, the optimized output distribution scheme can be obtained.

8. The method according to claim 7, wherein It also includes: Obtaining the heat source priority data, extracting game parameters from the data, the game parameters include cost and priority, calculating the output weight of each heat source in combination with the cost function, determining the output distribution, analyzing the output relationship between heat sources, if the output of a certain heat source exceeds the preset threshold, then adjusting the weight of the heat source to obtain the corrected output distribution, calculating the game result, determining the output balance state between heat sources, and adjusting the output scheme according to the priority through the sorting method, specifically including: Calculating the initial weight coefficient of the heat source according to the heat source operation cost data, and generating a heat source priority ranking table; Constructing an initial heat source output distribution matrix according to the heat source priority ranking table, and predicting the heat source output distribution data through a recurrent neural network; Performing threshold comparison on the heat source output distribution data, and reallocating the weight coefficient by using a linear weighted calculator to obtain the heat source output correction data; Using the heat source output correction data to construct a heat source competition relationship matrix, and calculating the game balance solution to obtain the output scheme.

9. The method according to claim 1, characterized in that According to the optimized output allocation scheme, summarize the total heat source emissions. If the total emissions exceed the preset threshold, reduce the output proportion of high-emission heat sources through the iterative adjustment algorithm, and output the target output allocation matrix, including: Obtain pollutant concentration data, and calculate the total heat source emissions dataset from the pollutant concentration data through an emission factor calculator; Perform a comparison operation on the total heat source emissions dataset and the preset emission limit, and generate heat source emission over-standard data through a concentration threshold comparator; Perform a descending order operation on the heat source emission over-standard data, and generate a heat source adjustment coefficient through an output adjustment calculator; Perform an output reduction operation on the heat source according to the heat source adjustment coefficient, limit the reduction step through the response rate constraint, and optimize the output adjustment scheme through an iterative calculator to generate a heat source output adjustment sequence; Establish a target output allocation matrix according to the heat source output adjustment sequence.

10. The method according to claim 1, characterized in that Identify the grid interaction attributes corresponding to the output allocation matrix, and identify the heat source output volatility. If the volatility is lower than the preset threshold, transmit the output allocation matrix to the heat source control system to determine the heat source output scheduling instruction, including: Read the heat source output data from the output allocation matrix, perform time-frequency analysis on the heat source output data through the wavelet decomposition algorithm, obtain the frequency fluctuation characteristics according to the grid interaction feature set, and calculate the heat source output volatility; If the heat source output volatility is lower than the preset volatility threshold, read the control reference parameters from the preset parameter library and generate a scheduling control parameter set; Dynamically predict the scheduling control parameter set using a recursive neural network, calculate the heat source response duration according to the heat source response characteristic curve, generate a heat source scheduling instruction, and send the heat source scheduling instruction to the heat source control device through a communication status detector.

Images